> Infection rate is orthogonal to preventing people from being infected. If you're not exposed, you can never be infected.
Sure, there are confounds: people who choose to be vaccinated don't behave exactly the same way that people who don't choose to be vaccinated do. Therefore, the observational data is somewhat confounded.
Two directions for the confound: people who are vaccinated may choose to take more risk after vaccination. People who are vaccinated may be those who take COVID seriously to begin with and be biased to take less risk in general. Probably the latter "wins", though I certainly do a lot more than I would do if I wasn't vaccinated.
However, IMO this is all moot: we have vaccines that showed massive efficacy in randomized, controlled trials (over a period of time), and case control studies that show ongoing high efficacy against Delta... and the confounded observational data shows a truly massive effect.
It also depends how you define efficacy/effectiveness. Does it mean for each exposure, over a period of time, etc.